Olives of Cornicabra variety were harvested during the 2012/13 olive season in the Agricultural Experimental Station “La Chimenea”, Instituto Madrileño de Investigación y Desarrollo Rural, Agrario y Alimentario (IMIDRA), located in the Autonomous Community of Madrid.

About 35 kg of olives were hand-picked in perfect sanitary conditions from olive trees at two ripeness stages: November 19th (VOO1) and December 31st (VOO2). The olive ripeness index was determined according to the method of Uceda and Frias (1975), based on the evaluation of the olive skin and pulp colors. Seven maturity states of the fruit were used in the evaluation: 0, bright-green skin; 1, green-yellowish skin; 2, green skin with reddish spots; 3, reddish-brown skin; 4, black skin with white flesh; 5, black skin with <50% purple flesh; 6, black skin with ≥50% and >100% purple flesh; and 7, black skin and purple flesh.

Olive oils were extracted within a 24-h period using the Abencor system (MC2 Ingenierías y Sistemas, Sevilla, Spain). The olives were crushed with a hammer mill, the paste was mixed at 26 ± 1 ºC for 30 min and then centrifuged at 3,500 rpm for 1 min. The oil was separated by decantation, filtered and stored at 4 ºC in the dark using amber glass bottles prior to analysis.

Domestic deep-fat fryers with a 1.1 L stainless steel vessel (SOLAC, Vitoria-Gasteiz, Spain) were used for potato frying. Spunta variety potatoes (Valencia, Spain) were fried. The initial surface-to-volume ratio of the oil was 0.20 cm^-1 (225 cm²/1100 cm³). The ratio between the amount of potatoes and the volume of frying oil was kept at 183 g/L.1 L during 40 repeated frying operations by replenishing every five frying operations with unused oil in order to maintain, insofar as possible, a constant oil-to-food ratio. The oils were heated for 10 min till 180 ºC and left for 8 min. The potatoes were cut into 2 mm thick slices using a domestic potato slicer, introduced into the oil at 180 ºC and fried for 6 min. The oil was left to cool to 30-35 ºC between same-day frying operations. The cooling time was approximately 4 h between frying operations. On the basis of a previous study (Bastida and Sánchez-Muniz, 2001), a total of 40 frying operations at the rate of 4 fryings per day were performed with each VOO. The total frying time and the entire operation time per day were 24 min and 13 h, respectively. The whole procedure was performed in duplicate using two fryers for each VOO. Thirty-five mL oil from each of the two fryers were taken after 10, 20, 30 and 40 frying operations and kept frozen at -20 ºC until analysis.

The analysis of fatty acid composition of the oils was performed by GC after derivatization to fatty acid methyl esters (FAME) with 2M KOH in methanol at room temperature (IUPAC, 1992).

The absolute amount of individual FAME was calculated by multiplying the percentage area of each fatty acid methyl ester by the amount of the unaltered FAME fraction expressed as parts per unit giving equivalent results to those obtained using an internal standard (Olivero-David et al., 2014).

Total PC in the fresh oils and after being used in frying were determined by adsorption chromatography (Dobarganes et al., 2000). One gram of oil was separated in a silica-packed chromatography column using 150 mL of hexane:diethyl ether (90:10, v/v) and 150 mL of diethyl ether to elute the non-polar and polar fractions, respectively. The amount of the non-polar fraction was determined gravimetrically and that of the polar fraction by weight difference.

To obtain further information about changes due to thermoxidation and hydrolysis during frying, an HPSEC analysis of the polar fraction was performed following a slight modification of the IUPAC method (Dobarganes et al., 2000). A solution of the polar fraction in tetrahydrofuran (10-15 mg/mL) was analyzed in a High-Performance Liquid Chromatograph (HPLC) (Agilent 1100 series, Madrid, Spain) equipped with a 20-μL loop, two 300 mm x 7.5 mm i.d. (5 μm particle size), 0.01 and 0.05 μm PL gel columns (Agilent, Madrid, Spain), connected in series, operating at 40 ºC, and a refractive index detector (Agilent Technologies 1260 infinity, Madrid; Spain). HPLC-grade tetrahydrofuran was used as the mobile phase with a flow rate of 1 mL/min. Triglyceride oligomers (TGO), triglyceride dimers (TGD), oxidized triglycerides (OTG), diglycerides (DG), monoglycerides (MG) and free fatty acids (FFA) were quantified in the PC fraction. Hydrolytic compounds (HC) were calculated as the sum of DG, MG and FFA, while thermoxidation compounds (TC) were calculated as the sum of TGO, TGD and OTG. Polymers were calculated as the sum of TGD and TGO.

Samples of the unused and used frying oils were saponified with 0.5M NaOH in ethanol by applying reflux heating during 10 min. Then methylation was performed according to the AOAC (1995) using 20% BF₃ in methanol with reflux heating during 15 min. After methylation, 1 g of sample was separated by adsorption chromatography on silica gel using 150 mL of hexane/diethyl ether (88:12, v/v) and 150 mL of diethyl ether to elute the non-polar (unaltered) and polar (altered) fractions of FAME, respectively (Márquez-Ruiz et al., 1995). As outlined above for the non-polar and polar oil fractions, the amounts of the unaltered and altered FAME fractions were also determined gravimetrically.

Both FAME fractions were analyzed by HPSEC following the same analysis described above for the oil polar fraction. Thermal fatty acid dimers (thermal-FAD) and non-oxidized fatty acid monomers (non-oxFAM) were quantified in the non-polar fraction, whereas fatty acid oligomers (FAO), oxidized fatty acid dimers (oxFAD), and oxidized fatty acid methyl esters (oxFAM) were quantified in the polar fraction (Márquez-Ruiz et al., 1995). Aliquots of both fractions (10-15 mg/mL tetrahydrofuran) were analyzed in the same HPSEC device and using the same conditions described above for the oil polar fraction.

Total phenolic compounds in unused and used frying oils were determined after methanol extraction, subsequent reaction with Folin-Ciocalteu reagent and spectrophotometrically determination at an absorption wavelength of 725 nm (Vázquez-Roncero et al., 1975).

The determination of α- and γ-tocopherols present in unused and used frying oils was carried out by high performance liquid chromatography according to the IUPAC method (1992).

The unused and used frying oils were analyzed by the Rancimat test to determine the OSI according to AOCS Official Method Cd 12b-92 (Firestone, 1998). A 743 Rancimat device (Metrohm Ltd, Herisau, Switzerland) was used. The OSI was obtained at 100 ºC with a continuous air flow of 20 L/h and using 2.5 g oil.

Unless it is indicated, all determinations were made in triplicate. The Pearson product-moment correlation test was used to find relationships between parameter data and alteration markers in the oils. Linear regressions were performed to ascertain linear adjustments between the concentration of PC, different TC and HC. The SPSS 19.0 statistical program was employed. Comparisons between linear adjustments for the different compounds in the two VOOs were checked by the ANCOVA test using SAS 9.2 statistical program. Statistical significance was set at p < 0.05.

Following conventional Madrid-Area harvesting, Cornicabra olives were harvested at 25 and 31 weeks after flowering, presenting ripeness indices of 2.08 and 4.13 for VOO1 and VOO2, respectively. These values were relatively lower compared to those reported for other olive varieties, e.g. Picual, harvested at the same season period (Olivero-David et al., 2014).

The oils presented very similar fatty acid compositions (Table 1). As expected, the levels of linoleic and linolenic acids were relatively low. The fatty acid composition covered the expected normal range for Cornicabra virgin oils, i.e. high levels of oleic acid and low levels of linoleic and stearic acids (Salvador et al., 2001a; b). This differs from VOOs obtained from other varieties (Olivero-David et al., 2014; Sánchez-Casas et al., 2003). Cornicabra is one of the Spanish virgin oils with the lowest amount of linoleic acid (Alba-Mendoza, 1996). Beltrán et al. (2005) suggested that the fatty acid composition of VOOs is quantitatively affected by two main factors, the olive variety and the ripeness stage. In the present study, despite the differences in the ripeness indices no substantial differences were found in the fatty acid compositions.

The PC levels were low and of the same order in the two oils, although slightly lower for VOO1 (1.90 g/100 g oil) compared to VOO2 (2.70 g/100 g oil). The values obtained reflected the good quality of both oils, as the PC levels for high-quality fresh oils normally range between 0.4 and 6.4 g/100 g oil (Lumley, 1988). The PC levels of the oils agreed with previous data obtained for extra VOOs (Olivero-David et al., 2014; Romero et al., 1995). Similarly to other high-quality olive oils (Bastida and Sanchez-Muniz, 2001), DGs were the major compounds in the polar fraction, followed by OTG. VOO1 showed a lower amount of OTG (0.39 g/100 g oil) than VOO2 (0.61 g/100 g oil). The oils did not differ substantially in the total HC. As expected, TGO and TGD were not detected. These are characteristic of oils that have been subjected to high temperatures like refined oils or used frying oils (Dobarganes et al., 1988).

The levels of total polyphenols were within those reported for Cornicabra virgin oils (Salvador et al., 2001a). VOO1 showed a remarkably greater content (273.6 mg/kg oil) than VOO2 (61.2 mg/kg oil). The amount of tocopherol, which was mainly constituted by α-tocopherol and also by γ-tocopherol in trace amounts, was also higher in VOO1 (174 mg/kg oil) than in VOO2 (128 mg/kg oil). Therefore, the total amount of antioxidants, i.e. polyphenols and tocopherols, was higher in VOO1.

Both oils presented high resistance to oxidative degradation, showing OSI values greater than 20 h at 100 ºC. Despite the greater amount of antioxidants found in VOO1 and the fact that no substantial differences were found in the fatty acid compositions between the two oils, the oxidative stability of VOO1 (OSI_100ºC 24.4 h) was comparable to that of VOO2 (OSI_100ºC 25.4 h). These results reflect the complexity of lipid oxidation, which depends not only on the fatty acid composition and levels of antioxidants, but also on a number of other factors (Velasco and Dobarganes, 2002).

Overall, the two oils studied with different ripeness indices presented similar fatty acid compositions, high quality and high and similar oxidative stability, but differed in the total amount of antioxidants, i.e. total polyphenols and tocopherols.

As expected, decreases in the amount of oleic (5.9% and 6.4%), linoleic (29.0% and 29.4%) and linolenic (42.9% and 28.6%) acids were observed in VOO1 and VOO2 after 40 frying operations, respectively. Oleic acid displayed the major decrease in absolute amounts (4.8-5.0 g/100 g of oil). Due to the great differences in the contents of oleic and linoleic acids in olive oils, significant decreases in oleic acid were not detected until linoleic acid was substantially reduced (Romero et al., 1995; 2000). The changes in oleic, linoleic and linolenic acid concentrations were negatively and linearly correlated (p < 0.05; p < 0.001; p < 0.001, respectively) with the number of frying operations (data not shown), which supports studies reported by Romero et al. (1995; 2000).

Figures 1 and 2 show the total PC and the TC and HC amounts, respectively. The PC and TC increased linearly (at least r² = 0.981; p < 0.001) with the number of frying operations, whereas, in agreement with previous studies (Bastida and Sánchez-Muniz, 2001; Romero et al., 2003; Velasco et al., 2005), the HC remained constant.

The levels of total PC indicated that the two oils showed similar frying performance. No substantial differences in the total level of alteration were found between the two oils during frying (Figure 1). These results differ from those reported for Picual VOOs, which were subjected to the same experimental frying conditions of the present study (Olivero-David et al., 2014). The oil of more mature olives exhibited quicker formation of PC, showing values that were greater in 3.5-4 g/100 g oil compared to two oils with lower ripeness index values. While no substantial differences in the fatty acid compositions were found among the three oils, the oil with the lowest ripeness index showed higher antioxidant levels, i.e. total polyphenols and tocopherols (Olivero-David et al., 2014).

The PC and TGP levels found at the end of the frying assay were not greater than 18 g/100 g and 8 g/100 g oil, respectively, which are far from the limits of 25 g/100 g oil for total PC and 10–12 g/100 g oil for TGP (DGF, 2000). This is indicative of the high frying life of Cornicabra virgin oils. The values of PC agreed with those of Casal et al. (2010) in commercial extra VOO of the same cultivar. This showed the highest stability in the frying of potatoes when compared to a mixture of refined oil and VOO, and refined sunflower oil.

Previous studies suggested that oil thermoxidation changes fit to linear adjustments when frying was performed with low or null oil-turnover but to a power, logarithmic or quadratic adjustment when frequent turnover was done (Sánchez-Muniz et al., 2008). According to the linear total PC level adjustments obtained (Table 2), both VOOs would have been discarded between the 55th and 59^th frying operations, respectively. This is relevant taking into account the relatively low turnover of fresh oil performed.

Results of polar FAME are given in Figure 3. From a nutritional point of view, the evaluation of polar and non-polar fatty acyl chains of triglycerides provides valuable additional information about oil degradation (Sánchez-Muniz et al., 2008). Unlike altered triglycerides which comprise both modified and unaltered fatty chains, the analysis of polar and non-polar FAME enables us to know the quantities of the alteration compounds that are directly absorbed after digestion. During frying, the increase in altered fatty acyl chains was linearly adjusted with the number of frying operations (at least p < 0.01), whereas the non-polar fraction and the non-oxFAM presented negative linear adjustments (all, p < 0.01) (Table 3). Previously, it was suggested that 27.6% PC corresponds to 8.7 to 11.3% polar-ME (Dobarganes and Márquez-Ruiz, 2007). Using 9.5% Polar-ME as a cut-off point, the VOOs shelf-life could have been extended at least to the 64^th frying.

Figure 4 shows the levels of total polyphenols and tocopherols. The concentrations of total polyphenols sharply decreased during the first 10 frying operations in the two oils, reaching undetectable levels after the 25^th frying operation. The amount of tocopherols also sharply decreased during frying, showing losses of approximately 90% in both oils after the first 5 frying operations. The tocopherol levels were undetected in the two oils after 10 frying operations.

The results in this study show the rapid degradation of natural antioxidants present in oils at the temperatures normally applied in the frying of foods. In this respect, the similar frying performance of the two oils, mainly differing in the contents of antioxidants, could be related to the rapid degradation of both polyphenols and tocopherols at frying conditions. Barrera-Arellano et al. (2002) suggested that the degradation of tocopherols at frying temperatures seems to be predominant over their antioxidative role, as the degradation of tocopherols added to different purified vegetable oils was independent of the degree of unsaturation of the oil. These authors suggested that the protection effect of naturally occurring tocopherols was a secondary mode of action.

As expected, the OSI values decreased progressively during frying as a consequence of the losses in antioxidants and the accumulation of oxidation compounds (Figure 5). No significant differences were found in the OSI values between the two oils throughout the frying assay. The determination of the oxidative stability of oils by the Rancimat test is a widely used method to determine the resistance to oxidation at the conditions of the test, demanding short analysis time. The oxidative stability index provides information about the oxidative behaviour of oils at room or moderate temperatures if oxygen is not a limiting factor. On the other hand, this index does not allow us to get to know the performance of an oil at the high temperatures applied in the frying of foods because of the different reaction mechanisms involved at frying conditions (Velasco and Dobarganes, 2002). The Rancimat results for the starting oils were similar, even though both oils showed quite different amounts of antioxidants. In contrast, Yousfi et al. (2006) found that changes in oil stability and phenolic compounds in olive oils during fruit ripening strongly differed according to the variety and the maturity level of the fruit. The Cornicabra virgin oils of the present study showed higher initial oxidative stability than the Picual virgin oils tested in a previous study (Olivero-David et al., 2014). This may be attributed in part to the lower amount of linoleic acid and higher content in oleic acid in the Cornicabra oils.